hsong@nvuxl.UUCP (g hugh song) (12/28/90)
Hi, I am having a terrible time wrestling with my program. It had run successfully on Cray, Stardent, IBM RS/6000, Sun, etc. Now I am trying to run on my machine---DecStation 5000/200. It ran on DS 5000/200, well, sort of. It is running with -O2 -C. However it does not run without -C. Usually the revese is the usual case. So I tried to use the debugger "dbx". (for the first time, since I always used one withe a graphical interface so far with Apollos) BTW, When will dec provide a graphical debugger for f77 in addition to cc? Alas, the source is not available although I compiled that routine with "-g". That's fine. I can open a separate window, "vi" the file with ":set number" I finally found that the bug could be found if there were a mechanism which finds floating point exception "NaN"---yeah! The famous "NaN"---which is *allowed* to be *undetected* by ANSI FORTRAN 77 Standard. Am I correct? It seems that the MIPS f77 compiler does not have floating-point-exception detection mechanism. in accordance with ANSI FORTRAN 77 Standard. Arrrrgggg.... I do not understand why ANSI did not specify such mechanism. If you know why, please enlighten me. Does MIPS Fortran compiler really not have such detection mechanism at all? The compiler version is 2.1. (cc uses 2.0). You know, version cc -O2 has a problem compiling TeX3.0. I do not think it has been solved with 2.1 which is used by f77 with Ultrix 4.0. So, what is the version number which comes with Ultrix version 4.1? My site does not have Ultrix 4.1, yet. Thanks. -hsong- nvuxl!hsong@bellcore.bellcore.com PS: You need to type "nvuxl\!hsong@..." on Csh command line.
fkittred@bbn.com (Fletcher Kittredge) (12/28/90)
In article <764@nvuxl.UUCP> hsong@nvuxl.UUCP (g hugh song) writes: >Hi, >I am having a terrible time wrestling with my program. >It had run successfully on Cray, Stardent, IBM RS/6000, Sun, etc. Now I am >trying to run on my machine---DecStation 5000/200. It ran on DS 5000/200, >well, sort of. I am having a difficult time understanding your message. Could you tell me if you linked with the regular math library (-lm) or the system V math library (-lmV)? If you were using the default library, you may want to relink using the system V math library, which seems to function better in the case of overflow/underflow/NaN errors. regards, fletcher